OUAHMANE TARIQ¶
1: UNDESTAND THE PROBLEM STATEMENT AND BUSINESS CASE
2: IMPORT LIBRARIES/DATASETS AND PERFORM EXPLORATORY DATA ANALYSIS
In [7]:
# import key libraries
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
from wordcloud import WordCloud, STOPWORDS
import nltk
import re
from nltk.stem import PorterStemmer, WordNetLemmatizer
from nltk.corpus import stopwords
from nltk.tokenize import word_tokenize, sent_tokenize
import gensim
from gensim.utils import simple_preprocess
from gensim.parsing.preprocessing import STOPWORDS
import plotly.express as px
# Tensorflow
import tensorflow as tf
from tensorflow.keras.preprocessing.text import one_hot,Tokenizer
from tensorflow.keras.preprocessing.sequence import pad_sequences
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense, Flatten, Embedding, Input, LSTM, Conv1D, MaxPool1D, Bidirectional, Dropout
from tensorflow.keras.models import Model
from tensorflow.keras.utils import to_categorical
In [8]:
# load the stock news data
stock_df = pd.read_csv("stock_sentiment.csv")
In [9]:
# Let's view the dataset
stock_df
Out[9]:
In [10]:
# dataframe information
stock_df.info()
In [11]:
# check for null values
stock_df.isnull().sum()
Out[11]:
3: PERFORM DATA CLEANING (REMOVE PUNCTUATIONS FROM TEXT)¶
In [12]:
import string
string.punctuation
Out[12]:
In [13]:
Test = '$I love AI & Machine learning!!'
Test_punc_removed = [char for char in Test if char not in string.punctuation]
Test_punc_removed_join = ''.join(Test_punc_removed)
Test_punc_removed_join
Out[13]:
In [14]:
# Let's define a function to remove punctuations
def remove_punc(message):
Test_punc_removed = [char for char in message if char not in string.punctuation]
Test_punc_removed_join = ''.join(Test_punc_removed)
return Test_punc_removed_join
In [15]:
# Let's remove punctuations from our dataset
stock_df['Text Without Punctuation'] = stock_df['Text'].apply(remove_punc)
In [16]:
stock_df
Out[16]:
In [17]:
stock_df['Text'][2]
Out[17]:
In [18]:
stock_df['Text Without Punctuation'][2]
Out[18]:
4: PERFORM DATA CLEANING (REMOVE STOPWORDS)¶
In [19]:
# download stopwords
nltk.download("stopwords")
stopwords.words('english')
Out[19]:
In [20]:
# Obtain additional stopwords from nltk
from nltk.corpus import stopwords
stop_words = stopwords.words('english')
stop_words.extend(['from', 'subject', 're', 'edu', 'use','will','aap','co','day','user','stock','today','week','year'])
In [21]:
# Remove stopwords and remove short words (less than 2 characters)
def preprocess(text):
result = []
for token in gensim.utils.simple_preprocess(text): #put all the words compose the text in a list
if len(token) >= 3 and token not in stop_words:
result.append(token)
return result
In [22]:
# apply pre-processing to the text column
stock_df['Text Without Punc & Stopwords'] = stock_df['Text Without Punctuation'].apply(preprocess)
In [23]:
stock_df['Text'][0]
Out[23]:
In [24]:
stock_df['Text Without Punc & Stopwords'][0]
Out[24]:
In [25]:
stock_df
Out[25]:
5: PLOT WORDCLOUD¶
In [26]:
# join the words into a string
stock_df['Text Without Punc & Stopwords Joined'] = stock_df['Text Without Punc & Stopwords'].apply(lambda x: " ".join(x))
In [27]:
# plot the word cloud for text with positive sentiment
plt.figure(figsize = (20, 20))
wc = WordCloud(max_words = 1000 , width = 1600 , height = 800).generate(" ".join(stock_df[stock_df['Sentiment'] == 1]['Text Without Punc & Stopwords Joined']))
plt.imshow(wc, interpolation = 'bilinear');
6: VISUALIZE CLEANED DATASETS¶
In [28]:
stock_df
Out[28]:
In [29]:
nltk.download('punkt')
Out[29]:
In [30]:
# word_tokenize is used to break up a string into words
print(stock_df['Text Without Punc & Stopwords Joined'][0])
print(nltk.word_tokenize(stock_df['Text Without Punc & Stopwords Joined'][0]))
In [31]:
# Obtain the maximum length of data in the document
# This will be later used when word embeddings are generated
maxlen = -1
for doc in stock_df['Text Without Punc & Stopwords Joined']:
tokens = nltk.word_tokenize(doc)
if(maxlen < len(tokens)):
maxlen = len(tokens)
print("The maximum number of words in any document is:", maxlen)
In [32]:
tweets_length = [ len(nltk.word_tokenize(x)) for x in stock_df['Text Without Punc & Stopwords Joined'] ]
tweets_length
Out[32]:
In [33]:
# Plot the distribution for the number of words in a text
fig = px.histogram(x = tweets_length, nbins = 50)
fig.show()
7: PREPARE THE DATA BY TOKENIZING AND PADDING¶
In [34]:
stock_df
Out[34]:
In [35]:
# Obtain the total words present in the dataset
list_of_words = []
for i in stock_df['Text Without Punc & Stopwords']:
for j in i:
list_of_words.append(j)
In [36]:
# Obtain the total number of unique words
total_words = len(list(set(list_of_words)))
total_words
Out[36]:
In [37]:
# split the data into test and train
X = stock_df['Text Without Punc & Stopwords']
y = stock_df['Sentiment']
from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.1)
In [38]:
X_train.shape
Out[38]:
In [39]:
X_test.shape
Out[39]:
In [40]:
X_train
Out[40]:
In [41]:
# Create a tokenizer to tokenize the words and create sequences of tokenized words
tokenizer = Tokenizer(num_words = total_words)
tokenizer.fit_on_texts(X_train)
# Training data
train_sequences = tokenizer.texts_to_sequences(X_train)
# Testing data
test_sequences = tokenizer.texts_to_sequences(X_test)
In [42]:
print("The encoding for document\n", X_train[1],"\n is: ", train_sequences[1])
In [43]:
# Add padding to training and testing to make all samples with same size
padded_train = pad_sequences(train_sequences, maxlen = 29, padding = 'post', truncating = 'post')
padded_test = pad_sequences(test_sequences, maxlen = 29, truncating = 'post')
In [44]:
#for examples of 3 samples we add 0 to have same size
for i, doc in enumerate(padded_train[:3]):
print("The padded encoding for document:", i+1," is:", doc)
In [45]:
# Convert the data to categorical 2D representation
y_train_cat = to_categorical(y_train, 2)
y_test_cat = to_categorical(y_test, 2)
In [46]:
y_train_cat.shape
Out[46]:
In [47]:
y_test_cat.shape
Out[47]:
In [48]:
y_train_cat
Out[48]:
8: UNDERSTAND THE THEORY AND INTUITION BEHIND RECURRENT NEURAL NETWORKS AND LONG SHORT TERM MEMORY NETWORKS (LSTM)¶
9: BUILD A CUSTOM-BASED DEEP NEURAL NETWORK TO PERFORM SENTIMENT ANALYSIS¶
In [49]:
# Sequential Model
model = Sequential()
# embedding layer
model.add(Embedding(total_words, output_dim = 512))
# Bi-Directional RNN and LSTM
model.add(LSTM(256))
# Dense layers
model.add(Dense(128, activation = 'relu'))
model.add(Dropout(0.3))
model.add(Dense(2,activation = 'softmax'))
model.compile(optimizer = 'adam', loss = 'categorical_crossentropy', metrics = ['acc'])
model.summary()
In [ ]:
# train the model
model.fit(padded_train, y_train_cat, batch_size = 32, validation_split = 0.2, epochs = 2)
10: ASSESS TRAINED MODEL PERFORMANCE¶
In [ ]:
# make prediction
pred = model.predict(padded_test)
In [ ]:
np.argmax?
In [ ]:
# make prediction
prediction = []
for i in pred:
prediction.append(np.argmax(i))
In [ ]:
# list containing original values
original = []
for i in y_test_cat:
original.append(np.argmax(i))
In [ ]:
# acuracy score on text data
from sklearn.metrics import accuracy_score
accuracy = accuracy_score(original, prediction)
accuracy
In [ ]:
# Plot the confusion matrix
from sklearn.metrics import confusion_matrix
cm = confusion_matrix(original, prediction)
sns.heatmap(cm, annot = True)